At a time when traditional news media struggle to engage younger audiences, we face a crucial challenge: How can we make complex political information accessible and relevant in a world where fewer people turn to conventional news sources with traditional — or even outdated — formats?

Aftonbladet’s solution was to experiment with a technology and format that has taken the world by storm: chatbots.

Using chatbots in a news context

We decided to test the concept on a major news event rather than creating a general chatbot for all our content (although that’s in the pipeline as well!) and chose the EU election.

This had several advantages:

• A vast amount of data was available: how the election works, the structure of the EU, party affiliations, candidate lists, and policy documents.
• As election day approached, a continuous stream of news would feed into the bot.

But there were also risks:

• Building a chatbot requires significant time and investment.
• The EU election, at least in Sweden, is not known for high engagement levels.
• Politics is a sensitive topic. What if the chatbot gave incorrect answers? What if someone asked, “Which party should I vote for?” and a screenshot of the bot’s response ended up on social media?
• Most importantly, journalists are used to having complete control — every letter and number we publish is meticulously reviewed, and everything falls under our editorial responsibility.

Despite these challenges, our leadership team approved the idea. And as it turned out, they made the right decision.

The birth of the Election Buddy 

The result was the Election Buddy, an unexpected success story that combined the best of both worlds: our journalism and new technology. It became so successful (more on that later) that we launched a follow-up chatbot for the U.S. election, The U.S. Election Buddy.

Our Election Buddies are AI-driven chatbots that allow users to ask questions about elections and politics and receive immediate, fact-based answers.

Aftonbladet, with its tabloid background and close relationship with readers, wanted the bot to take on a character-like presence, offering friendly, informal, and even humorous responses.

Unlike generative AI solutions like ChatGPT, our bots are built on a knowledge database curated and fact-checked by Aftonbladet’s journalists and editors. Every piece of information is carefully selected and verified, eliminating the risk of misinformation or biased responses — an essential safeguard in today’s information landscape.

A new way to consume news

Instead of scrolling through lengthy articles, users receive direct answers to their questions in less than a second. For those who find traditional news formats challenging, the chatbot provides summaries or explains complex concepts, making political news more accessible to a broader audience.

Election Buddies put users in control of their news experience. They can explore topics of interest, ask follow-up questions for deeper understanding, or request simplified explanations of complex issues.

Simply put: The experience starts from their existing knowledge level.

How we built our buddies

The success of our Election Buddies rests on a carefully developed approach that combines journalistic integrity with technological innovation:

• A journalism-based database. We created a comprehensive knowledge base consisting of thousands of sources, all derived from articles and surveys produced by Aftonbladet’s journalists. This ensured the chatbot’s answers were always accurate and aligned with our editorial standards. The database was continuously updated with fresh data, and post-election, we integrated results to keep the conversation relevant.
• Clear limitations. Unlike many open AI systems, our bots had well-defined boundaries. They could only answer election and political questions, relying solely on verified information from our database.
• User testing and iterative improvement. We worked closely with test groups to understand what types of questions users wanted to ask and which responses were most useful. This iterative approach allowed us to refine prompts and response quality over time.
• A cross-functional team. The project brought together editorial expertise, AI specialists, and UX designers. This interdisciplinary approach ensured a balance between journalistic quality and technical innovation.

Results exceed expectations

Despite generally low engagement levels with the EU election in Sweden, our chatbot quickly gained traction:

• EU Election Buddy answered 150,000 questions about EU politics in just a few weeks.
• 78,000 users found it valuable enough to log in.
• U.S. Election Buddy was an even bigger success, answering 600,000+ questions on election day alone.
• More than 25% of users were under 36 years old, a demographic that traditional news media often struggle to reach.

Beyond the numbers, the chatbot also provided unexpected benefits for our newsroom. During the 2020 U.S. election, Aftonbladet received 9,000 reader questions but could only answer about 300.

With Election Buddy, editors in 2024 could redirect simpler, fact-based questions to the bot, freeing up resources to tackle more complex issues. Additionally, we gained valuable insight into what our audience wanted to know during election periods.

5 key takeaways

The success of Election Buddies has provided us with several key lessons:

1. AI doesn’t replace journalism, it enhances it. The chatbot made political news more accessible and adaptable without compromising journalistic integrity.

2. Today’s media consumers expect content tailored to their needs. AI tools like Election Buddies enable personalised news experiences that traditional publishing cannot.

3. Context matters. We helped readers engage with the bot by integrating it into articles and providing pre-prompted questions. A generic “Ask anything about politics” prompt likely wouldn’t have generated as much interaction.

4. Good AI tools take time. Building a robust knowledge base and fine-tuning responses took months of work.

5. Users may be skeptical of AI-generated content but will engage if it provides value. If you ask readers whether they want AI-driven news, they may say no. But if you give them an excellent service, they will use it.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .